Overview / Usage

Why segregate?

• 47% of the waste going into landfills is organic waste and 18% is recyclable waste.
• At present, 68% of the total waste that Chennai produces everyday comes from households.
• In an ideal situation, we can reduce this** at least by half.**
• If our residences, commercial establishments and institutions simply managed organic and recyclable waste properly, we can keep over 60% of our waste out of landfills!
• Furthermore, the problem is that the people who want to help don’t often know how to segregate the wastes.
• Our app aims to do just that. Not only will we be classifying the objects, we will be educating them on what can be recycled as well.

What change do we aim to Establish?

• 40% of all paper, 18% of all plastic and 60% of all glass waste produced by the city is currently handled by the informal sector.
• This means that they currently keep approximately 35% of all paper, plastic and glass waste generated out of our landfills! Although significant, this is only 1/3rd of all the recyclable waste generated.
• By simply segregating our recyclables from our organics, we can boost the revenue three-fold and keep an additional 21,829 tonnes of waste out of the landfill every month!
• 47% of the total waste going to landfills is organic waste. At the household level, about 60% of our waste is organic. If we composted our organic waste, that constitutes the largest chunk of waste being kept out of landfills.

​

Methodology / Approach

The Process

1. We collected over 5000 images from the internet and classified them into 3 Classes : Recyclable, Non Recyclable and Compost.
2. We designed a simple CNN for object detection and classification.
3. The Neural Network was trained on the 5000 images following a 80:20 train:test split, giving a 90% accuracy.
4. The model was saved and converted to a .tflite file to use on Android Application.
5. The Android App was built and the model was mounted on the app. The App was perfected and converted to an apk.

​

The Neural Net

_________________________________________________________________

Layer (type) Output Shape Param #

=================================================================

conv2d_1 (Conv2D) (None, 222, 222, 32) 896

_________________________________________________________________

max_pooling2d_ (None, 111, 111, 32) 0

_________________________________________________________________

conv2d_2 (Conv2D) (None, 109, 109, 64) 18496

_________________________________________________________________

max_pooling2d_2 (None, 54, 54, 64) 0

_________________________________________________________________

flatten_1 (Flatten) (None, 186624) 0

_________________________________________________________________

dense_1 (Dense) (None, 128) 23888000

_________________________________________________________________

dropout_1 (Dropout) (None, 128) 0

_________________________________________________________________

dense_2 (Dense) (None, 3) 387

=================================================================

Total params: 23,907,779

Trainable params: 23,907,779

Non-trainable params: 0

_________________________________________________________________

​

User Experience

Our process is simple :

1. Open trashify.
2. Point to waste.
3. Find out which bin to put the waste in.
4. Trash the waste as a responsible citizen.

​

​

Technologies Used

Tensorflow Lite

Keras

Android Studio

h5py

Repository

https://github.com/RohitMidha23/trashify